Achieve Seamless HTTP Autoscaling with Kedify

by Zbynek Roubalik, Founder & CTO

July 25, 2024

At Kedify, we’re committed to pushing the boundaries of cloud-native Kubernetes autoscaling. Today, we’re excited to introduce our new production-ready HTTP Scaler for KEDA, designed to deliver unmatched performance, stability, and ease of use for HTTP-based autoscaling in Kubernetes environments. You can now use Kedify to scale up from or down to zero using HTTP.

Why Choose Kedify’s HTTP Scaler Over the Upstream KEDA HTTP Add-On?

Production Readiness: The upstream KEDA HTTP Add-On is still in its beta phase, which means it may lack the robustness required for production environments. In contrast, Kedify’s HTTP Scaler is fully production-ready, offering reliable and stable performance that you can count on for your mission-critical applications.

Enhanced Stability and Performance: Built on top of the upstream HTTP Add-On, Kedify’s HTTP Scaler incorporates significant improvements to enhance stability and performance. Our enhancements ensure that your applications can handle high traffic loads efficiently without compromising on reliability.

Elimination of Single Points of Failure: A critical drawback of the upstream HTTP Add-On is its interceptor component, which can become a single point of failure. Kedify’s solution mitigates this risk by integrating the industry-proven Envoy proxy, ensuring enhanced reliability and eliminating single points of failure.

Simplified Configuration: Kedify’s HTTP Scaler uses the existing ScaledObject Custom Resource Definition for configuring HTTP-based scaling, making the process straightforward and user-friendly. In contrast, the existing upstream HTTP Add-On relies on the more complex HTTPScaledObject in addition to KEDA’s existing ScaledObject configuration, which can complicate configuration and management. This adds complexity especially within GitOps environments where maintaining configuration consistency and simplicity is important.

Enterprise-Grade Support: Receive support from experts with extensive experience in deploying HTTP-based scaling across diverse use cases and cloud-native environments.

Key Features of Kedify’s HTTP Scaler

Automatic Network Traffic Wiring: Kedify’s HTTP Scaler simplifies the configuration process by automating the necessary Ingress reconfigurations for scaling. Unlike the upstream add-on, which requires users to manually handle these configurations, Kedify’s solution automates this process, saving time and reducing the potential for errors.

Fallback Mechanism: In case of issues within the scaling control plane, Kedify’s HTTP Scaler includes an automatic fallback mechanism that reroutes traffic to ensure continuous availability. This feature ensures that your applications remain accessible even during control plane disruptions and automatically resumes scaling once the control plane is stable.

Built-In Health Checks: Kedify’s HTTP Scaler is equipped with built-in health checks, ensuring that applications scaled to zero can still respond with proper health statuses. This prevents the application from appearing as failed in your monitoring stack, providing a more accurate representation of your system’s health.

Performance and Envoy Integration: Kedify’s HTTP Scaler leverages the power of the Envoy proxy for superior performance and reliability. By integrating Envoy, Kedify ensures high performance and robust traffic management. Additionally, if you already have Envoy set up in your environment, Kedify’s HTTP scaler will work seamlessly with existing configurations, offering a smooth and efficient integration.

Performance Metrics

To showcase the performance benefits of Kedify’s HTTP Scaler, we deployed a simple web application that displays an HTML page and responds with an artificial delay of 0.3 seconds. The application was deployed to an AKS cluster with 4 nodes. We used the hey tool to generate load with the following command:

hey -z 5m -c 1500 http://app-url

Below is a comparison table between the upstream KEDA HTTP Add-On, Kedify’s HTTP Scaler, and a scenario without autoscaling. The resource usage metrics represent the combined usage of KEDA + HTTP Scaler and all its components:

Metric	Kedify HTTP Scaler	KEDA HTTP Add-On	No Autoscaling	Description
Avg Response Time	0.5184s	0.7959s	0.5075s	The average time taken to respond to requests. Lower is better.
Total CPU Usage	115m	2198m	51m	Total CPU consumption in millicores (m). 1000m = 1 core. For example, 115m = 11.5% of a core.
Total Memory Usage	308Mi	547Mi	167Mi	Total memory usage in mebibytes (Mi). Lower usage is preferred.
Request Efficiency	High	Moderate	None/Low	Efficiency in handling requests under varying loads.

Key Takeaways

These metrics highlight Kedify’s superior performance and resource efficiency by reducing the average response time by approximately 35% compared to the upstream HTTP Add-On. This improvement, along with more efficient resource usage, makes it an ideal choice for production environments. It’s important to note that these data reflect the scaling of a single application; resource usage will increase as more applications are scaled.

ScaledObject Configuration

For this benchmark, we used the following ScaledObject configuration of the Kedify HTTP Scaler, for the upstream we configured a similar HTTPScaledObject.

kind: ScaledObject
apiVersion: keda.sh/v1alpha1
metadata:
  name: http-demo-scaledobject
spec:
  scaleTargetRef:
    name: http-demo-deployment
  cooldownPeriod: 5
  minReplicaCount: 0
  maxReplicaCount: 10
  triggers:
    - type: kedify-http
      metadata:
        hosts: demo.keda
        service: http-demo-service
        port: '8080'
        scalingMetric: requestRate
        targetValue: '10'

How to Get Started

Getting started with Kedify’s HTTP Scaler is easy. Follow the step-by-step guide in our Quickstart documentation to install and configure the HTTP Scaler. For detailed information on how to use and configure the new HTTP Scaler, please refer to the HTTP Scaler documentation.

For an overview of the HTTP Scaler’s features and benefits, check out our HTTP Scaler page.

Reduce cost and enhance performance.

Start autoscaling with Kedify today.

Get Started Free

Conclusion

Kedify’s HTTP Scaler offers a robust, production-ready solution for scaling based on HTTP traffic in Kubernetes environments. It offers better performance, stability, and ease of use compared to the upstream KEDA HTTP Add-On, making it an excellent choice for cloud-native applications. By automating network setup, integrating the Envoy proxy, and simplifying configuration, Kedify makes scaling using HTTP effortless.

We invite you to try out Kedify’s HTTP Scaler and experience the benefits firsthand. For any questions or support, please contact us at info@kedify.io.

Kedify Elastic Autoscaling Datasheet